How many tokens is my prompt, and what does it cost?
Tokens are not words. Estimating them wrong is how a prompt gets silently truncated, or how an API bill comes in at four times what you expected.
The short answer
Computed from this tool’s default settings — characters in your prompt and the rest as most people start. Change them below for your own case.
About 2,105 tokens at 3.8 characters each. Held in context on a 8B model that costs 0.36 GB of KV cache — charged in memory locally, and per request on a hosted API.
Prompt Token Counter
Estimate token count from characters, and the memory and money it uses.
About 2,105 tokens at 3.8 characters each. Held in context on a 8B model that costs 0.36 GB of KV cache — charged in memory locally, and per request on a hosted API.
What each setting changes
Every input moves the result for a reason. This is what each one does and where to find the value for your own machine.
| Setting | Default | What it changes |
|---|---|---|
| Characters in your prompt | 8000 | Paste your text elsewhere and use its character count. |
| Characters per token | 38 | English prose averages about 3.8. Code and non-Latin scripts are denser. |
| Model size | 8B — Llama 3.1 8B class | Capacity follows total parameters, even for mixture-of-experts models. |
Real answers across model size
The same calculation run at a range of settings, with everything else left at its default. These are computed by the tool itself, not written by hand.
| Model size | Estimated tokens | KV cache for this | At $10 / M input | At $0.07 / M input |
|---|---|---|---|---|
| 1B — Llama 3.2 1B class | 2,105 | 0.13 GB | $0.021 | $0.0001 |
| 7B — Mistral / Qwen 7B class | 2,105 | 0.33 GB | $0.021 | $0.0001 |
| 14B — Qwen 14B class | 2,105 | 0.47 GB | $0.021 | $0.0001 |
| 32B — Qwen 32B class | 2,105 | 0.71 GB | $0.021 | $0.0001 |
| 120B — gpt-oss 120B class | 2,105 | 1.38 GB | $0.021 | $0.0001 |
About 2,105 tokens at 3.8 characters each. Held in context on a 1B model that costs 0.13 GB of KV cache — charged in memory locally, and per request on a hosted API.
How this is calculated
No lookup tables and no invented constants. Here is the arithmetic, so you can check it against your own numbers.
Token counts depend on the tokeniser, so this is an estimate rather than an exact figure. English prose averages roughly 3.8 characters per token; code, JSON and non-Latin scripts produce more tokens for the same characters.
Worth knowing when migrating between models: the tokeniser introduced with Claude Opus 4.7 produces roughly 30% more tokens for the same text than older models, so a cost estimate built on an earlier model under-counts.
These are well-founded engineering estimates, not benchmark results. Your quantisation, runtime and context length all move the real number, and usable memory is an assumption rather than a specification. See the full methodology for every assumption behind these figures.
Who this is for
The three situations that bring people to this calculation.
Fitting a prompt
Check a document fits before sending it.
Budgeting
Turn a corpus into a cost.
Chunking
Size RAG chunks in the unit models actually use.
How to use this calculator
Four steps, no account, nothing leaves your browser.
Set characters in your prompt
Start at the top of the panel. Every figure recalculates as you change it — there is no submit button, because watching the number move is the point.
Adjust the rest to match your setup
2 further settings: characters per token, model size. Defaults are the common case, so change only what differs for you.
Read the headline, then the table
The large figure answers the question. The table underneath shows how the answer changes across nearby settings, which is usually where the decision actually gets made.
Check it against the method
The arithmetic is written out above. If a number looks wrong for your hardware, the assumptions are the first place to look — usable memory and quantisation are the two that vary most.
How many tokens is my prompt, and what does it cost: common questions
The questions people ask about this, answered without hedging.
How many tokens is a word?
English prose runs about 0.75 words per token, or roughly four characters per token. Code is denser at around 3.2–3.6 characters, because punctuation and indentation each consume tokens.
Do different models tokenise differently?
Yes. Each family ships its own tokeniser, and counts can differ by 10–20% between them for the same text. Treat any estimate as approximate unless you run the real tokeniser.
Why does my prompt get cut off?
The context window covers input and output together. If the input nearly fills it, there is no room left to answer, and most runtimes drop tokens from the front rather than warning you.
Do non-English languages use more tokens?
Usually yes, often substantially. Tokenisers are trained mostly on English, so other scripts fragment into more pieces per word.
Related local AI tools
All 50 run on the same arithmetic, so answers across them agree.
More on context
Will it fit
Speed
Quantisation
Runtime & errors
Cost
Fine-tuning
Searches this page answers
Different ways of asking the same question, all resolved above.
Now find the models that fit
Sizing is only half the problem. Model Radar takes your hardware and shows which models actually run on it, ranked by what they are good at — the same arithmetic as this page, applied to every model worth running.